The data for the article: Explainable machine learning classification of Chandra X-ray sources: SHAP analysis of multi-wavelength features

This repository provides the complete data for Table 1 in the paper titled : "Explainable machine learning classification of Chandra X-ray sources: SHAP analysis of multi-wavelength features" (submitted to ApJ, under review). This dataset contains the classification probabilities (Class membership probabilities) for 269366 point sources given in the Chandra Source Catalogue. The class membership probabilities are assigned using LightGBM model trained on 7703 sources with known classes. The sources are classified in the following 8 class of X-ray emitting astrophysical sources (class label given in the data-table is mentioned in the brackets): Active Galactic Nuclei (AGN) Young Stellar Object (YSO) X-ray emitting Stars (STAR) High Mass X-ray Binary (HMXB) Low Mass X-ray Binary (LMXB) Cataclysmic Variable (CV) Ultra-luminuous X-ray Sources (ULX) Pulsar (PULSAR) For training and prediction we used multiwavelength data from Gaia, 2MASS, SDSS, WISE and GALEX in addition to the properties from the CSC-2.0 master catalog.Table description:Columns: NAME (Observation ID of the source in the CSC-2.0) RA (J2000) in deg Dec (J2000) in deg class 1: Predicted class with highest CMP CMP1: probability for highest probable class class 2: Predicted class with second highest CMP CMP2: probability of second highest class. The data is providedd in two formats: an original comma-separated version and a version formatted to match the specifications of the AAS Journals and CDS. The latter version was supplied by one of the AAS data editors. For an interactive visualisation of the data-table see the github repository chandra-XAI. The repository provides the local and global SHAP explanation to these classification. These explanations are obtained using one-vs-rest classifier for the majority classes: AGN, Stars and YSO.